Description of problem:

After running [1] as part of [2], all the computes have a nova_compute container that is either unhealthy or bouncing in a loop (varies). This occurs in a Juniper Contrail backed environment .

The errors observed are mainly along the lines of [3] which, according to Juniper,are due to the bindmounts for the Nova container pointing towards /usr/lib/python3.6/site-packages (which do not exist yet) rather than the previous /usr/lib/python2.7/site-packages/ .


If the upgrade  is continued on a given compute through the LEAP part and beyond, the nova containers become functional and healthy . 

We have the following questions in this context :

1) Is the observed behavior normal/expected (i.e. is it safe to continue FFU in spite of this) ? 
2) If not, what is the expected behavior and what is the actual impact of observed behavior versus expected behavior (if any) ?  
3) If behavior is normal/expected should it be documented ? 
4) Is this possibly Contrail specific or likely to affect other third-party ML2 setups ? 



[1]
openstack overcloud upgrade run [--stack <stack_name>] --playbook upgrade_steps_playbook.yaml --tags nova_hybrid_state --limit all


[2]
https://access.redhat.com/documentation/en-us/red_hat_openstack_platform/16.2/html-single/framework_for_upgrades_13_to_16.2/index

[3]
ERROR oslo_service.service nova.exception.InternalError: Failure running os_vif plugin plug method: No VIF plugin was found with the name vrouter


Version-Release number of selected component (if applicable):


How reproducible:
Undetermined


Steps to Reproduce:
1. Run a 13-->16.2 FFU up to step "openstack overcloud upgrade run [--stack <stack_name>] --playbook upgrade_steps_playbook.yaml --tags nova_hybrid_state --limit all" on a Juniper Contrail backed ENV


Actual results:
Observe unhealthy or bouncing in a loop (varies) nova_compute containers on all computes


Expected results:
Unclear (BZ created with aim to answer that question and others)



Additional info: